Shopify licenses Liquid AI's foundation models
Liquid AI has signed a multi-year, multi-million-dollar deal to put its small foundation models into Shopify's search and recommendations, starting with a text model that completes a search in under 20 milliseconds.
Liquid AI announced a multi-year partnership under which Shopify will license Liquid's foundation models (LFMs) for search and other uses across its platform. Liquid describes the agreement as worth multiple millions of dollars but has not disclosed the financial terms.
The first production deployment is a text model that improves Shopify search and completes a query in under 20ms. The two companies also built a generative recommender system together, using what Liquid calls a novel HSTU architecture. According to Liquid, the recommender beat Shopify's previous system in controlled testing and produced higher conversion rates from recommendations. The company did not publish the test figures.
Liquid says its LFMs are built for multimodal inference under 20ms without giving up quality. It claims that on specific production-like tasks, LFMs with about 50% fewer parameters beat open-source models including Qwen3, Gemma3 and Llama 3, while running inference 2–10× faster. These are the company's own results and have not been independently verified.
The deal covers a license for LFMs across Shopify workloads that need low latency and high quality, along with ongoing joint R&D and a shared roadmap. The companies are evaluating multimodal models for customer profiles, agents and product classification.
The agreement makes official the co-development work the two companies were already doing. It follows Shopify's participation in Liquid AI's $250 million Series A round in December 2024. Shopify CTO Mikhail Parakhin and Liquid AI CEO Ramin Hasani discussed the partnership in a joint interview.